UK MoD JSP 936 Dependable AI in Defence Part 1: Directive (V1.1, 2024) 69-71: Bias and harm mitigation: analyse data, algorithms and models for unwanted bias, and monitor and mitigate harms including from intentional bias

What does 69-71 require?

Data, learning algorithms and models are examined for unwanted bias capable of causing unintended harm, which include physical and psychological harm and discrimination against protected characteristics. Wherever harm could result, strategies to monitor and mitigate it are developed with sufficient understanding of the AI behaviour, and even deliberate bias has its ethical harms weighed and mitigated.